Skip to main content

Table 1 Classification results on a toy network

From: Framework for role discovery using transfer learning

  Micro-F1 Macro-F1
RolX (Target tuned) 0.757 0.537
struc2vec (Target tuned) 0.865 0.731
Proposed method (Target tuned) 0.894 0.730
Proposed method (Model selection) 0.880 0.699
  1. Target tuned means the model selected on the target labels, which is the upper bound performance each model can achieve, and model selection means the model selected with the validation network